Abstract: The main objective of this study is to simulate the motion of one million heavy particles in turbulence using a point particle model and one-way coupling, based on high-precision compressible homogenous isotropic turbulence direct numerical simulation data. This work focuses on two aspects. Firstly, it uses spectral truncation filters with different filter widths to obtain large-scale flow fields and investigates the influence of different filter scales on particle motion. Secondly, this study sets five different initial particle velocities to investigate the evolution of particle clustering and kinematic properties. In the study of particle clustering, Shannon entropy is used to describe the instantaneous clustering of particles, while statistical results in the steady state are described using probability density distribution functions. The results show that the filter scale has different effects on particle clustering for different Stokes numbers. Specifically, small-scale flow structures promote particle clustering for low Stokes numbers, while they inhibit clustering for high Stokes numbers. Additionally, with increasing Stokes numbers and decreasing truncation wavenumbers, the probability density distribution of particle velocity and acceleration becomes more concentrated. Furthermore, this work also finds that differences in initial particle velocities have a significant impact in the early stages of evolution but eventually converge to the same statistical steady state. This discovery emphasizes the complexity of particle motion in turbulence and the importance of statistical characteristics.